The study cohort comprised 8,936 adults free of diabetes at baseline. Follow-up spanned a median 9.20 years, during which 1,038 participants developed type 2 diabetes. Investigators measured 71 metabolites in paired plasma and urine using capillary electrophoresis–mass spectrometry. They then constructed separate biospecimen-specific metabolomic risk scores (MRSs) with a fivefold cross-fitting approach to limit overfitting. Cox proportional hazards models adjusted for lifestyle variables, a polygenic risk score, body-mass index (BMI), fasting plasma glucose, and HbA1c analyzed associations with incident diabetes.
After adjustment for clinical covariates and mutual inclusion of both biospecimen scores, the urine MRS remained associated with incident diabetes. The hazard ratio per standard deviation was 1.20 with a bootstrap 95% confidence interval of 1.07–1.25. The plasma MRS did not show a similar independent association (HR per SD 1.02; 95% CI 0.91–1.18). Data indicated that urine-derived metabolic variation carried predictive information that was not captured by the paired plasma panel and standard clinical risk markers.
When investigators evaluated metabolite-level associations and controlled the false discovery rate, 20 urine metabolites remained significant. The source did not list each metabolite in the abstract; the full preprint and supplementary materials provide detailed identities and coefficients. No plasma metabolites survived false discovery rate adjustment in the same analyses. These findings indicate that specific urinary metabolites, measured in this 71-metabolite panel, were more robustly tied to later diabetes risk than their circulating counterparts in this cohort.
The study team compared coefficients between plasma and urine for overlapping metabolites. Coefficients differed significantly for betaine and threonate. That divergence suggests some metabolites may reflect distinct physiological or excretory pathways when measured in urine versus plasma. Investigators reported no plasma metabolites with statistically significant associations after FDR correction, whereas multiple urine metabolites remained associated, pointing to specimen-specific signal rather than a uniform systemic signature.
Urine-MRS associations persisted in subgroup analyses restricted to participants with lower levels of selected risk markers. The study team tested for interaction by baseline glycaemia, BMI, polygenic risk score, and by plasma MRS; interaction p-values exceeded 0.2 for these tests, indicating no strong evidence that these factors modified the urine-MRS effect. The authors used bootstrap confidence intervals in several analyses; they also applied FDR procedures at the metabolite level. The source did not report additional model diagnostics or incremental discrimination metrics in the abstract; those details are available in the full preprint supplementary files.
This report is a preprint and has not undergone peer review. Investigators declared no competing interests. Ethical approval was provided by the Medical Ethics Committee of Keio University School of Medicine (approval number 20110264); all participants provided written informed consent. Individual deidentified participant data are not publicly posted because of privacy and ethical restrictions. According to the authors, deidentified data required to reproduce the analyses will be made available upon reasonable request after peer-reviewed publication, subject to TMCS and ethics committee approval and a data-use agreement. Funding sources included the Japan Society for the Promotion of Science, Japan Science and Technology Agency (JST-OPERA) with Konica Minolta, Yamagata Prefectural Government, Tsuruoka City, and Konica Minolta, Inc.
Limitations explicitly noted in the preprint include its prepublication status and the observational design; the abstract does not list other specific study constraints such as detailed model calibration, potential residual confounding, or generalizability beyond the Japanese population. The full manuscript and supplementary material should be consulted for comprehensive methods, the complete list of significant urine metabolites, and additional sensitivity analyses.